Google offers enhanced generative AI features for cloud customers

By Davey Alba

Alphabet Inc’s Google announced a slate of upgraded artificial intelligence features for its cloud-computing clients as the technology giant tries to catch up with rivals, including the allied forces of Microsoft Corp. and startup OpenAI, who have taken advantage of the AI boom.
 

The company on Wednesday unveiled Gemini Pro for enterprises, allowing developers to build applications using Google’s latest AI model, which was announced last week. Gemini is a large-scale AI system trained on vast amounts of data that can generate new content based on what users request. 

Google Cloud clients can use Gemini to create apps such as AI-powered chatbots, easy-to-query inventory databases and marketing presentations. The company also emphasized that Gemini Pro will be free at launch for cloud customers, with some limits. Ultimately, Google said, it plans to ensure that its cloud AI offering will be “competitively priced.” The company said Gemini Pro’s text-based capabilities are four times less expensive for input and two times less expensive for output than the last iteration of its AI model, PaLM 2, which was released in June.

 

The AI model is built to “generalize and seamlessly understand, operate across, and combine different types of information, including text, code, audio, image, and video in the same way that humans see, hear, read, listen and talk about different types of information simultaneously,” Google Cloud Chief Executive Officer Thomas Kurian said.

Though Google is a pioneer of generative AI, its products have lagged behind in popularity. Last week, Google aimed to counter that narrative with the release of Gemini, which has three sizes: Ultra, Pro and Nano. The company has begun rolling out Nano, the smallest version, which runs directly on devices like Google’s flagship smartphone, the Pixel 8 Pro. It has also released a specially designed version of Gemini Pro in Bard, Google’s AI chatbot and answer to OpenAI’s wildly popular app, ChatGPT. With the release of Gemini Pro to app developers and businesses, Google wants to send the message that it is no longer trailing OpenAI and the startup’s latest AI system, GPT-4, which is available through partner Microsoft’s Azure cloud-computing platform.

Google said Gemini Pro supports 38 languages across 180 countries worldwide, and currently accepts text as input and generates text as output. Google is also releasing a dedicated Gemini Pro Vision platform that can handle text- and image-based prompts from users. 

Gemini Ultra, the company’s largest and most capable model for “highly complex” tasks, will be available to select cloud customers and partners for early experimentation before it’s released to the general public next year, the company said.

Google announced that Gemini Pro would be incorporated into two key cloud products: Google AI Studio and Vertex AI. Google AI Studio, a free, web-based developer tool, is what Google called the “fastest way to build with Gemini.” The tool allows clients to use the Gemini API for developing apps. Vertex AI, meanwhile, provides developers and cloud clients more customization. Businesses will be able to customize Gemini using their company’s own data, and build Gemini-powered search tools and chatbots, among other applications.

Kurian said Gemini Pro’s pricing is getting “significantly more attractive.” Developers will have free access to Gemini Pro and Gemini Pro Vision through Google AI Studio, the company said, which is suitable for most app development needs. Vertex AI, which is more flexible, will be free until early next year.

Google also announced upgraded versions of previously released AI models. Imagen 2, Google’s text-to-image technology, will have improved photorealism, text rendering and logo generation capabilities, the company said. Google is unveiling MedLM, too — a family of models fine-tuned for the health care industry, building on the work the company has done on Med-PaLM 2, Google’s AI model that has been trained on specialized medical knowledge.

The company also announced a global partnership with Mistral AI, the buzzy, Paris-based artificial intelligence startup with an emphasis on open-source software. Mistral AI will distribute some AI products, including optimized proprietary language models, on Google Cloud’s infrastructure, the two companies said in a blog post.

Though the agreement is non-exclusive, Mistral AI’s CEO and co-founder, Arthur Mensch, praised Google Cloud for its flexible tools and its ability to support the company’s products. “Google Cloud’s open source support and responsible AI principles, extensive and reliable infrastructure capabilities, and commitments around privacy and security align strongly with our mission to develop openly available models,” Mensch said.